Compression encoding of video typically includes intra or inter prediction to reduce spatial and temporal redundancies in the video. The prediction results, i.e., prediction residuals representing differences between video pixel values and predicted pixel values, may be transformed into coefficients in the frequency domain. The coefficients are referred to as transform coefficients and may be large in number. The transform coefficients are quantized into quantization values, and the quantization values are entropy encoded. The larger the number of transform coefficients the higher the computational complexity encountered when the transform coefficients are quantized. Higher computational complexity disadvantageously results in lower overall video encoding throughput. Therefore, decreasing computational complexity when the number of transform coefficients is large is desirous, especially in demanding video applications that require high encoding throughput, such as real time video and multi-person gaming, for example.
In the drawings, the leftmost digit(s) of a reference number identifies the drawing in which the reference number first appears.